LaoWang 0a80a28cc5 feat: 新增 9 篇关键缺位论文 (189篇) ...

架构 5 篇:
- MiniMax-01: Lightning Attention + MoE (2501.08313)
- Byte Latent Transformer: tokenization-free (2412.09871)
- Large Concept Models: 句子级推理 (2412.08821)
- xLSTM: LSTM复兴, matrix memory (2405.04517)
- LLaDA: Diffusion LLM, 非自回归 (2502.09992)

后训练 2 篇:
- MiniMax-M1: CISPO RL 推理涌现 (2506.13585)
- Coconut: 潜在空间连续推理 (2412.06769)

Agent 1 篇:
- Generative Agents: Stanford AI Town (2304.03442)

多模态 1 篇:
- GPT-4V/4o/5.4: 视觉-语音原生多模态

2026-06-02 12:53:44 +00:00

..

papers.json

feat: 新增 9 篇关键缺位论文 (189篇)

2026-06-02 12:53:44 +00:00